Longwang: Zero-Shot Global Spatiotemporal Precipitation Downscaling with a Latent Generative Prior

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Wang, Yue, Visioni, Daniele
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866917506032074752
author Wang, Yue
Visioni, Daniele
author_facet Wang, Yue
Visioni, Daniele
contents High-resolution precipitation information is essential for climate impact assessment, yet global climate models remain too coarse to resolve key small-scale processes. Existing machine learning downscaling methods often require paired low- and high-resolution data for supervised learning, are tied to fixed regions or scale factors during inference, and can be computationally expensive to train and run in physical space. Here we introduce Longwang, a zero-shot latent generative framework for global spatiotemporal precipitation downscaling. Longwang learns a context-conditioned latent generative prior and combines it with a physically informed observation operator through posterior sampling, enabling daily O(10 km) precipitation fields to be generated from monthly O(100 km) inputs. On ERA5 reanalysis, Longwang outperforms standard posterior sampling with an unconditional generative prior in reconstructing fine-scale spatial patterns, preserving temporal coherence, and recovering extreme precipitation intensities. The framework further generalizes to historical climate simulations and future climate projections under substantial distribution shift.
format Preprint
id arxiv_https___arxiv_org_abs_2605_17603
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Longwang: Zero-Shot Global Spatiotemporal Precipitation Downscaling with a Latent Generative Prior
Wang, Yue
Visioni, Daniele
Atmospheric and Oceanic Physics
Machine Learning
High-resolution precipitation information is essential for climate impact assessment, yet global climate models remain too coarse to resolve key small-scale processes. Existing machine learning downscaling methods often require paired low- and high-resolution data for supervised learning, are tied to fixed regions or scale factors during inference, and can be computationally expensive to train and run in physical space. Here we introduce Longwang, a zero-shot latent generative framework for global spatiotemporal precipitation downscaling. Longwang learns a context-conditioned latent generative prior and combines it with a physically informed observation operator through posterior sampling, enabling daily O(10 km) precipitation fields to be generated from monthly O(100 km) inputs. On ERA5 reanalysis, Longwang outperforms standard posterior sampling with an unconditional generative prior in reconstructing fine-scale spatial patterns, preserving temporal coherence, and recovering extreme precipitation intensities. The framework further generalizes to historical climate simulations and future climate projections under substantial distribution shift.
title Longwang: Zero-Shot Global Spatiotemporal Precipitation Downscaling with a Latent Generative Prior
topic Atmospheric and Oceanic Physics
Machine Learning
url https://arxiv.org/abs/2605.17603